Making the Bet on Open Source
Making the Bet on Open Source
Using the tools before they were popular. Take a look at this reflection on Talend's decision to use Docker and Kubernetes while they were in their early phases.
Join the DZone community and get the full member experience.Join For Free
As a CTO today, if you received an email from your head of Engineering saying, "Can we say that Docker is enterprise production ready now?" your answer would undoubtedly be "yes." If you hadn't started leveraging Docker already, you would be eager to move on the technology that Amazon and so many other leading companies are now using as the basis of their applications' architectures. However, what would your reaction be if you had received that email four years ago when Docker was still far from stable, lacked integration, support, or tooling with all the major operating systems and Enterprise platforms, on-premise or cloud? Well, that is the situation that we at Talend were facing in 2015.
By sharing our approach and our learnings from choosing to develop with Docker and Kubernetes, I hope that we can help other CTOs and tech companies' leaders with their decisions to go all-in with today's emerging technologies.
Increasing Docker Use from Demos to Enterprise-Ready Products
Back in 2014, as we were architecting our next generation Cloud Integration platform, microservices and containerization were two trends that we closely monitored.
Talend, which is dedicated to monitoring emerging projects and technologies identified Docker as a very promising containerization technology that we could use to run our microservices. That same year, one of our pre-sales engineers had heard about Docker at a Big Data training and learned about its potential to accelerate the delivery of product demos to its prospects as a more efficient alternative to VMWare or Virtual Box images.
From that day, Docker usage across Talend has seen an explosive growth, from the pre-sales use case of packaging demos to providing reproduction environments to tech support or quality engineering and of course its main usage around service and application containerization for R&D and Talend Cloud.
During our evaluation, we did consider some basic things like we would with any up-and-coming open source technology. First, we needed to determine the state of the security features offered by Docker. Luckily, we found that we didn't need to build anything on top of what Docker already provided which was a huge plus for us.
Second, like many emerging open source technologies, Docker was not as mature as it is today, so it was still "buggy." Containers would sometimes fail without any clear explanation, which would mean that we would have to invest time to read through the logs to understand what went wrong — a reality that anyone who has worked with a new technology understands well. Additionally, we had to see how this emerging technology would fit with our existing work and product portfolio, and determine whether they would integrate well. In our case, we had to check how Docker would work with our Java-based applications & services and evaluate if the difficulties that we ran into there would be considered a blocker for future development.
Despite our initial challenges, we found Docker to be extremely valuable and promising as it greatly improved our development lifecycle by facilitating the rapid exchange and reuse of pieces of work between different teams. In the first year of evaluation, Docker quickly became the primary technology used by QA to rapidly setup testing environments at a fraction of the cost and with better performance compared to the more traditional Virtual environments (VMWare or VirtualBox).
After we successfully used Docker in our R&D processes, we knew we had made the right choice and that it was time to take it to the next level and package our own services for the benefit of our customers. With the support of containers and more specifically Docker by major cloud players such as AWS, Azure or Google, we had the market validation that we needed to completely "dockerize" our entire cloud-based platform, Talend Cloud.
While the choice to containerize our software with Docker was relatively straightforward, the choice to use Kubernetes to orchestrate those containers was not so evident at the start.
Talend's Road to Success with Kubernetes
In 2015, Talend started to consider technologies that might orchestrate containers which were starting to make up our underlying architectures, but the technology of choice wasn't clear. At this point, we faced a situation that every company has experienced: deciding what technology to work with and determining how to decide what technology would be the best fit.
At Talend, portability and agility are key concepts, and while Docker was clearly multiplatform, each of the cloud platform vendors had their own flavor of the orchestration layer.
We had to bet on an orchestration layer that would become the de facto standard or be compatible with major cloud players. Would it be Kubernetes, Apache Mesos, or Docker Swarm?
Initially, we were evaluating both Mesos and Kubernetes. Although Mesos was more stable than Kubernetes at the time and its offering was consistent with Talend's Big Data roadmap, we were drawn to the comprehensiveness of the Kubernetes applications. The fact that Google was behind Kubernetes gave us some reassurance around its scalability promises.
At the time, we were looking for container orchestration offerings, using Mesos required that we bundle several other applications for it to have the functionality we needed. On the other hand, Kubernetes' applications had everything we needed already bundled together. We also thought about our customers: we wanted to make sure we chose the solution that would be the easiest for them to configure and maintain. Last — but certainly not least — we looked at the activity of the Kubernetes community. We found it promising that many large companies were not only contributing to the project but were also creating standards for it as well. The comprehensive nature of Kubernetes and the vibrancy of its community led us to switch gears and go all-in with Kubernetes.
As with any emerging innovative technology, there are constant updates and project releases with Kubernetes, which results in several iterations of updates in our own applications. However, this was a very small concession to make to use such a promising technology.
Similar to our experience with Docker, I tend to believe that we made the right choice with Kubernetes. Its market adoption (AWS EKS, Azure AKS, OpenShift Kubernetes) proved us right. The technology has now been incorporated into several of our products, including one of our recent offerings, Data Streams.
Watching the team go from exploring a new technology to actually implementing it was a great learning experience that was both exciting and very rewarding.
Our Biggest Lessons in Working with Emerging Technologies
Because we have been working with and contributing to the open source community since we released our open source offering Talend Open Studio for Data Integration in 2006, we are no strangers to working with innovative, but broadly untested, technologies or sometimes uncharted territories. However, this experience with Docker and Kubernetes has emphasized some of the key lessons we have learned over the years working with emerging technologies:
- Keep your focus: During this process, we learned that working with a new, promising technology requires that you keep your end goals in mind at all times. Because the technologies we worked with are in constant flux, it could be easy to get distracted by any new features added to the open source projects. It is incredibly important to make sure that the purpose of working with a particular emerging technology remains clear so that development won't be derailed by new features that could be irrelevant to the end goal.
- Look hard at the community: It is incredibly important to look to the community of the project you choose to work with. Be sure to look at the roadmap and the vision of the project to make sure it aligns with your development (or product) vision. Also, pay attention to the way the community is run-you should be confident that it is run in a way that will allow the project to flourish.
- Invest the time to go deep into the technology: Betting on innovation is hard and does not work overnight. Even if it is buggy, dive into the technology because it can be worth it in the end. From personal experience, I know it can be a lot of work to debug but be sure to keep in mind that the technology's capabilities — and its community — will grow, allowing your products (and your company) to leverage the innovations that would be very time consuming, expensive and difficult to build on your own.
Since we first implemented Docker and Kubernetes, we have made a new bet on open source: Apache Beam. Will it be the next big thing like Docker and Kubernetes? There's no way to know at this point — but when you choose to lead with innovation, you can never take the risk-free, well-travelled path. Striving for innovation is a never-ending race, but I wouldn't have it any other way.
Published at DZone with permission of Laurent Bride , DZone MVB. See the original article here.
Opinions expressed by DZone contributors are their own.